Ioannidis' paper says 2.49%-4.16% of Santa Clara County has already been infected. Santa Clara County has a population of 1.9 million. That means we should expect (1900000 * 0.0249) * 0.0024 == 113 deaths at the lower bound, and (1900000 * 0.0416) * 0.0024 == 189 deaths at the upper bound. Santa Clara County has already recorded 134 deaths, so that number seem to fit. But it fits with the hard, categorical data from NYC only if we assume 100% of NYC to have already been infected.
TL;DR: Unless you're prepared to believe that
1) most of NYC residents have already been infected, or
2) that IFR is significantly greater than 0.24% and the actual number of COVID-19 deaths in Santa Clara County is multiples of what's reported
then it's clear that Ioannidis' paper is wrong. #2 seems far more plausible and may even be true, but the controversy is the implication of his paper on derived IFR. And if #2 is what's happening, the IFR is definitely greater than what people applying the paper (including, AFAIU, Ioannidis' himself) predict.
I am not sure why we think the IFR would be constant by region, though. Just doing an age banded calculation is enough to show that it is variable for populations. Such that it seems safe to assume other things could cause variation, as well.
An early thought I had was age being a proxy to lung damage from long term exposure to air pollution. This still holds for the population in NYC, as many of them are from there. Same for Italy. Many elderly are retired locals. Notable that both places have some of the worst air in the world. Contrast with California, which has been very progressive with emissions. I don't have data on age ranges in California. And I confess I have stopped following the news on this. (Peek in the occasional story that makes it here. Starting to pay more attention again.)